On the development of a Power Quality Benchmarking model

On the Development of a Power Quality Benchmarking Model Johan Rens School of Electrical and Electronic Engineering North-West University Potchefstroom, South Africa [email protected] Abstract—The development of a Power Quality (PQ) Benchmarking model is presented. Practical results are discussed to motivate the need of adding value to recorded PQ data by means of benchmarking PQ performance. The importance, science and methodology of cleaning up recorded sag and swell data per site and per region is demonstrated. Benchmarking of voltage distortion, unbalance and magnitude regulation is briefly introduced. Keywords-component; PQ benchmarking, voltage sags and swells, voltage unbalance, harmonic distortion, voltage regulation I. INTRODUCTION The quality of electrical energy is internationally recognised as more comprehensive than quality of service by means of reliability aspects. The economic impact of PQ on energy and business efficiency, environmental issues and others requires responsibility and accountability of all role players in the electrical industry. Standards on electrical power quality (PQ) that set compatibility and limit criteria are in wide-spread use (such as [1]). These standards describe PQ from a Quality of Supply (QoS – voltage) perspective and can include Quality of Use (QoU – current) aspects. The shared goal is to attain a minimum level in the quality of electrical energy throughout production, transmission, distribution and usage. It is important to note that most standards pertain to a single Point of Delivery (PoD) in the electrical network and contain guidelines on reporting quality aspects at a PoD. PQ data is nowadays readily available as a result of modern technology being applied in instrumentation and communication technology. It follows that visibility on parameters that quantify and 978-1-4244-5172-2/09/$26.00 ©2009 IEEE qualify the quality of electrical energy is straightforward. But, to understand how well an electrical utility (for example) is dealing with it's core business (electrical energy), an agreed upon methodology is needed. Scientific and other publications on how to develop, implement and apply a PQ benchmarking model is limited [2], [3], [4], [5], [6], [7]. The spirit of a PQ Management system is to use PQ data collected all over the electrical network to formulate management and other intervention practices to continually improve the quality of electrical energy served. A PQ Benchmarking model is needed to generate key performance parameters on all subsets of an electrical industry. PQ Benchmarking is a much more comprehensive concept than a PQ standard because a system perspective is now required. This paper deals with the experience gained in translating PQ Data to practical PQ information for the Namibian Electricity Control Board. Complicated PQ indices is of little practical value when it is important to empower operational personnel. II. PQ MANAGEMENT AT THE NAMIBIAN ELECTRICITY CONTROL BOARD The Namibian Electricity Control Board (ECB) supports it's regulatory role by collecting PQ Data from regional electricity distributors (RED's). The instruments records voltage based data which is then concentrated over a GSM network to a central server. Data collected over a two year period (2006 - 2008) was analysed and used in the first phase in development of a PQ Benchmarking model. The goal is to understand the level in quality of electrical energy by which these RED's serve users and to identify areas of improvement. Voltage sags and swells are probably the most studied aspect of PQ as the relation with reliability of service is well known. Various organisations have reported benchmarking efforts [5], [8], [9]. A. Recorded sag and swell data is not publishable data The difference in requirements to a proper perspective on sag an swell performance of both a single site and an area containing more than one site, is presented below. slight increase in voltage magnitude in the highest phase can cause the recording of voltage swells, 33 in this example. RMS profile of Voltages: Recording of swells 111 110 109 108 RMS III. III.VOLTAGE SAG AND SWELL BENCHMARKING 107 Va Vb Vc 106 Time (ms) Impact of Voltage Magnitude and Unbalance on recording of sags and swells: a per metering site analysis If an instrument is set up to record a voltage sag or well event in terms of the NRS 048-2007 definition thereof and the instrument was configured to detect RMS voltage variations around a nominal fixed value, than network operating conditions are likely to affect the number of sags and swells recorded. It is general practice to have such an agreement on a fixed nominal voltage level at a PoD in the network from a compatibility perspective. Steady-state network operating conditions can adversely affect the number of sags and swells recorded as shown below. Observe the example of voltage dips and swells recorded at a site in Namibia during a 2 month period shown in Figure 1. Figure 2: Operating voltage causing swells to be recorded The normal significance of a voltage swell is that a temporary rise in voltage magnitude (above 110%) has occurred for a short period of time (less than 3 seconds per NRS 048 definition) as a possible result of transient operations such as switching of loads and lightning. The high incidence rate seen in Table 1 are thus masking the true voltage swells character of this site. A similar argument could be filed for the recording of voltage sags when the operating voltage is set a rather low value that results in the recording of numerous dips when load operation cause a reduction in voltage level below the dip threshold, but which could be negligible in relative terms. The impact of network voltage imbalances are shown in Figure 3. It was found that 53 voltage dips can be attributed to the occurrence of this example of voltage imbalances at this PoD. 111 105 110 RMS profile of of Voltages: Recording of swells RMS profile unbalanced Voltages Figure 1: Voltage dips and swells recorded at a site in Namibia: July - Sept 06 The number and type of events are listed in Table 1. Data as recorded and upon filtering with a time criteria is shown. About 25 voltage events have been recorded per day when raw data is published as in Figure 1. It seems that a compatibility issue exists (more than 1 event per hour). TABLE 1: NUMBER AND TYPE OF VOLTAGE EVENTS PLOTTED IN FIGURE 1 Dip Type Dip type Y Dip type X1 Dip type X2 Dip type T Dip type S Recorded 2024 43 4 2 36 Filtered 175 20 2 1 6 Dip type Z1 Dip type Z2 Voltage Swells 79 2 65 1 1 23 The instruments used were configured to retain the profile of RMS voltages during an event. Analysis indicated a high voltage setting at this PoD shown in Figure 2. It is clear that a RMS RMS 109 100 108 95 107 !" !# 106 90 !$ Va Time (ms) 85 Vb Vc 80 Time (ms) Figure 3: Voltages in unbalance causing dips to be recorded The true voltage sag character of such site is masked when the impact of network steady state operation as shown above is not removed from recorded data. Publishing of all the sags and swells recorded add little value to Power Quality Benchmarking from a statistical perspective. Voltage regulation and voltage imbalance are network operational aspects that, if properly managed to compliancy and limit criteria, should not impact the recording of sags and swells. Network incidents that cause voltage sags and swells in a power system will rarely occur within milliseconds from each other (as in the examples above). Simple application of a time criteria upon populating a PQ database, will eradicate most of the impact of network operating conditions on sags and swells. A filter statement that ignore all voltage sags or swells within a 30 second interval but retaining the worst event only, was rigorously tested with the Namibian data. Not one example could be found of a typical root cause to a voltage sag such as lightning (eg. causing a single line to earth fault) and occurring within 30 seconds at the same site. development of a PQ Benchmarking model. NORED Voltage Dips for Aug 07 to Jul 08: UNFILTERED 1600 Z2 1400 Z1 1200 The filtered results in Table 1 retained the voltage incidents that were of “typical” causes and are now down to about 2.5 per day which is a total different perspective on compatibility at this site. This practical demonstration show that voltage sags and swells due to network operation can simply be flagged and removed by a time filter criteria. Take note that if the network condition such as a high operating voltage remain, a voltage sag will still be recorded every 30 seconds, but can be used as an alarming signal requesting manual intervention in data flagging in addition to the automated filtering proposed. The impact of voltage incidents on voltage events recorded: a per area analysis Voltage sag and swell data require additional analysis on a per area basis. Assume that an area is defined according to some geographical consideration and that various voltage levels exists in transmission and distribution networks in this area. Benchmarking of sags and swells with respect to the performance of an area requires distinction between PQ incidents and PQ events. A PQ event refers to any voltage event at a metering point whilst PQ incident refers to the root cause (such as lightning). One lightning strike in an area could result in numerous events recorded, but from a system perspective, only one incident occurred in the area. Time-stamping of PQ events at different metering sites due to the same incident can be different due to clock-drift. It is necessary to correctly group and identify incidents whilst accounting for difference in time-stamping. A similar approach is presented in [8]. Observe the graph in Figure 4 depicting all voltage dips (at various voltage levels) recorded in an area representing a Regional Electricity Distributor (RED) in Namibia. Limited information results. A superficial interpretation is that lightning coincides with the dip incidence rate as per typical weather pattern in Namibia. Accounting for clock-drift with a 60 second criteria results in the plot shown in Figure 5 which present a total different character in terms of lightning. But. the high occurrence of dip type Y shadows the other dip types. As the NRS 048 requires equipment to be at least compatible to dip type Y, removal thereof results in the plot shown in Figure 6. A more useful perspective is now revealed . Compatibility evaluation indicate that dip types S, Z1 and Z2 is of concern. Research of root causes to these dips could support formulation of intervention procedures that in future could reduce the occurrence. This is a simple demonstration on the importance of a proper perspective on voltage sags in the T 1000 S X2 800 X1 600 Y 400 200 Aug-07 Oct-07 Dec-07 Feb-08 0 Apr-08 Jun-08 Z2 Z1 T X1 Y S X2 Figure 4: Voltage sags as recorded for an area in Namibia: Unfiltered NORED Voltage Dips for Aug 07 to Jul 08: FILTERED 500 Z2 450 Z1 400 T 350 S 300 X2 250 X1 200 Y 150 100 50 Aug-07 Oct-07 Dec-07 Feb-08 0 Apr-08 Jun-08 Z2 Z1 T X1 S X2 Y Figure 5: Voltage sags for an area in Namibia: Filtered The impact of filtering per incident on the number of sags is lastly shown in Table 2 in final support of the statement that recorded sag data in area requires a system perspective in order to furnish a practical appreciation on the true sag character of that area. Histogram NORED Voltage Dips for Aug 07 to Jul 08: FILTERED 1400 120% 1200 100% 160 Z2 140 1000 80% 120 T 100 S Frequency Z1 800 60% 600 40% 400 80 20% 200 X2 60 0 Figure 6: Voltage sags for an area in Namibia: Filtered without type Y 571 299 Z1 359 310 T 310 171 S 1019 611 X2 429 289 X1 468 318 Y 10725 3111 13881 5109 Total IV. BENCHMARKING VOLTAGE WAVEFORM DISTORTION Non-sinusoidal waveform conditions increase commensurately with the growth in sophistication in energyconversion (and efficiency) due to versatility and controllability these solid-state technology offers and the latter has been seen to offer solutions at a fast rate of increase in power levels. Most of these solutions withdraw current in a non-linear fashion and although normally more energyefficient, some impact on the voltage waveform distortion could be expected as source impedances cannot be perfectly zero. Voltage waveform distortion when quantified by Voltage Total Harmonic Distortion (VTHD) is managed by compliancy levels (NRS 048 part 2 of 2007 set the VTHD level for networks below 33 kV to 8% and above to 4%). This section present a brief overview on benchmarking VTHD from a system perspective. A year-on-year benchmarking in VTHD was done by means of an annual histogram per site. Any PoD has to meet the compliancy criterium for 95% of the time (=95% of the acquired values). A typical example is shown in Figure 7 and Figure 8. 600 100% 80% 400 60% 300 40% 200 20% 100 0 0. 00 % Aug 07 – Jul 08: Filtered Frequency Aug 07 – Jul 08: Unfiltered 120% 500 TABLE 2: IMPACT OF FILTERING PER AREA Type Z2 4. 00 % Histogram 700 0% 4. 00 % T 3. 20 % Z1 2. 40 % Z2 1. 60 % Jun-08 Figure 7: Histogram of VHTD with CPF curve superimposed at a site in Namibia: 06/07 0. 80 % Apr-08 X1 S X2 3. 20 % VTHD 20 Aug-07 Oct-07 Dec-07 Feb-08 2. 40 % 1. 60 % 0% 0. 00 % 40 0. 80 % 0 X1 VTHD Figure 8: Histogram of VHTD with CPF curve superimposed at a site in Namibia: 07/08 Both sites are in compliance with the 95% CPF values of 2.6% and 3.8% respectively for the periods 06/07 and 07/08 although indicating an increase. Comparison of the distribution of VTHD values further indicate a different distribution with maximum values in an upward trend. Concentrating VTHD information recorded at multiple sites to system information is not straightforward. Application of the System Average Total Harmonic Distortion (SATHD) and System Average Excessive Total Harmonic Distortion (SAETHD) as formulated in [9] was not yet fully tested but seems to be a fair approach as it discounts for the size of the load where the VHTD was recorded against the total load being served by the section of the network this load is connected to. V. V.BENCHMARKING OF VOLTAGE UNBALANCE Application of the histogram on a per site basis was done for voltage unbalance (Vub) and is useful in benchmarking a site against itself similar to the previous section on VTHD. From a system perspective various approaches are possible [9]. The histograms in Figure 9 and Figure 10 are “statistics of statistics” but useful in visualising the distribution of Vub values in that area and in terms of year-on-year comparison. Histogram of Vub (95% CPF) for an area; 06/07 8 150% 6 100% 4 50% 2 2.0% 1.8% 1.6% 1.4% 1.2% 1.0% 0.8% 0.6% 0.4% 0.2% 0 Source impedance changed 0.0% Number 95% CPF A network configuration change (supply impedance change in September 06) can be seen in Figure 12 which caused a wider band of values during that time although during the rest of the period the network supply impedance proved to be sufficient, as can be deducted from the narrow spread of values around nominal. 0% Vub Figure 9: Histogram of 95% CPF values in Vub for an area in Namibia; 2006/2007 Histogram of Vub (95% CPF) for an area; 07/08 10 120% 80% 6 60% 4 40% 2.0% 1.8% 1.6% 1.4% 1.2% 1.0% 0.8% 0% 0.6% 0 0.4% 20% 0.2% 2 0.0% Number VII. CONCLUSION 100% 95% CPF 8 Figure 12: NRS 048 7 day sliding assessment of Voltage Magnitude at a sit in Namibia Vub Figure 10: Histogram of 95% CPF values in Vub for an area in Namibia; 2006/2007 VI. VI.BENCHMARKING OF VOLTAGE MAGNITUDE Voltage magnitude is one of the most important parameters in an ageing power system to optimise equipment availability. It requires daily benchmarking but historic information on a per site basis is useful in revealing network characteristics. A typical NRS 048 7-day sliding assessment for two different PoD's in the network is shown in Figure 11 and Figure 12. The quality of electrical energy is gaining widespread recognition as an important parameter in sustainable business processes. The development of PQ standards has matured and experience in application is growing. It is important to evaluate and report the success thereof. The science of PQ Benchmarking has to supplement PQ standards by research on how to extract information from PQ data and on how to present this information to support the core business of both utilities and the user industry. Experience in the development of a PQ benchmarking model reported in this paper indicate that more research and development is required and that more experience has to be gained to understand the user requirements of organisations that put a premium to the quality of the electrical energy in their daily business. REFERENCES 1] 2] 3] Tap changer setting changed 4] 5] 6] 7] Figure 11: NRS 048 7 day sliding assessment of Voltage Magnitude at a site in Namibia The voltage magnitude shown in Figure 11 is regulated within a narrow band (network supply impedance sufficient). The impact of a tap changer temporarily being set higher during August 2006 can be seen. 8] 9] NRS048 2007, South Africa. L.J. Honkapuro et. al., "Power Quality Factors in Efficiency Benchmarking," 18th International Conference in Electricity Distribution, Turin, 6-9 June 2005 M. McGranaghan, "Overview of International Voltage Quality Benchmarking Efforts," IEEE PES Transmission and Distribution Conference, Dallas, Texas, May 2006 EPRI TR-106294-V2, "An Assessment of Distribution System Power Quality. Vol. 2: Statistical Summary Report", Electric Power Research Institute, Palo Alto, Calif., May 1996. Daniel Brooks, Bill Howe, Establishing PQ Benchmarks, E Source, Boulder, Colo., May 2000. EPRI TR-107938, EPRI Reliability Benchmarking Methodology, EPRI, Palo Alto, Calif., 1997. D. Sabin, T. E. Grebe, M. F. McGranaghan, A. Sundaram, “Statistical Analysis of Voltage Dips and Interruptions—Final Results from the EPRI Distribution System Power Quality Monitoring Survey,” Proceedings 15th International Conference on Electricity Distribution (CIRED ’99), Nice, France, June 1999. Math H.J. Bollen, Understanding Power Quality Problems. IEEE Press, Piscataway, 2000. R. C. Dugan, M. F. McGranaghan, S. Santoso, H.W. Beaty, Electrical Power System Quality, McGraw-Hill, New York, 2nd edition, 2003